Files
coder/agent/agentcontext/manager_test.go
T
Kyle Carberry cd3692c0c2 feat: add agent-side workspace context sources and Agent API v2.10 PushContextState (#25983)
Adds the agent half of the workspace context sources RFC. The agent now
resolves instruction files, skills, and MCP configs into a typed
`Snapshot`, watches the relevant paths recursively, exposes the source
list over a workspace-agent HTTP API, and pushes each `Snapshot` to
coderd over a new `PushContextState` RPC on Agent API v2.10.

The coderd-side handler is a stub returning `Unimplemented` for now.
Real persistence to `workspace_agent_context`, chatd hydration on dirty
events, and the `KindMCPServer` MCP provider are tracked by
[CODAGT-569](https://linear.app/codercom/issue/CODAGT-569/enable-agent-api-v210-pushcontextstate-bump-currentminor-wire-coderd).
This matches the pattern used for v2.7 `ReportBoundaryLogs` in
[#21293](https://github.com/coder/coder/pull/21293), which bumped the
version and shipped a stub server so the wire and client could iterate
before the persistence layer landed.

## What ships

### agent/agentcontext (new package)

- `Source`, `Resource` (kinds `instruction_file`, `skill`, `mcp_config`,
`mcp_server` plus reserved `plugin`/`hook`/`subagent`/`command`),
`ResourceStatus`, `Snapshot`, `ComputeAggregateHash`.
- `Manager` owns the in-memory source list, performs the initial resolve
synchronously in `NewManager`, runs a re-resolve/watcher loop in `Run`,
exposes
`AddSource`/`RemoveSource`/`Sources`/`HasSource`/`Snapshot`/`SubscribeChanges`/`Resync`/`SeedSources`/`Close`.
- `Resolver` walks scan roots, classifies recognized files, enforces 64
KiB per-resource, 2 MiB aggregate, and 500-resource caps with
`StatusOversize`/`StatusExcluded`/`StatusUnreadable`/`StatusInvalid`
outcomes, skips `node_modules`/`vendor`/etc., validates symlink targets
stay inside the scan root, stamps `SourcePath` on user-derived
resources, and optionally pulls MCP server tool lists via an
`MCPProvider` interface. MCP config resources ship metadata only (size,
hash) so secrets in env blocks never leave the agent.
- `Watcher` is a recursive `fsnotify` wrapper with a 250 ms debounce,
dynamic arming of newly created directories, and an ENOSPC-tolerant
degraded mode that no-ops further syncs until the manager resyncs
explicitly.
- HTTP API for `GET/POST /sources`, `GET/DELETE /sources/{path}`, `POST
/resync` mounted at `/api/v0/context`.
- `Pusher` interface plus `RunPush` goroutine with exponential backoff
capped at 30 s. `DRPCPusher` adapts the generated `DRPCAgentClient210`
to `Pusher` and translates `drpcerr.Unimplemented` to
`ErrPushUnimplemented` so the push loop exits cleanly when talking to
coderd deployments that have not enabled the real handler.

### agent/proto (v2.10)

- New messages `ContextResource`, `PushContextStateRequest`,
`PushContextStateResponse` and the `PushContextState` RPC on `service
Agent`.
- Generated `DRPCAgentClient210` interface and
`codersdk/agentsdk.Client.ConnectRPC210` / `ConnectRPC210WithRole`.
- `tailnet/proto.CurrentMinor` bumped from `9` to `10`.

### Agent wiring

- `agent.Options.Client` declares both v2.9 and v2.10 connectors;
`run()` dials with `ConnectRPC210WithRole`.
- `apiConnRoutineManager` holds a `DRPCAgentClient210`. Existing v2.8
routines keep their narrower `DRPCAgentClient28` signature thanks to
interface embedding.
- `startAgentAPI210` is the v2.10 counterpart to `startAgentAPI` for
routines that need the new client. The push context state routine uses
it.
- A `contextManager` is constructed in `agent.init()`, seeded from the
existing `CODER_AGENT_EXP_*_DIRS` env vars, started in its own goroutine
under `gracefulCtx`, and closed in `agent.Close`.
- `handleManifest` calls `Manager.SeedSources` for sources rooted at the
manifest directory, then `Resync` after `manifest.Swap`, so the snapshot
reflects the workspace working directory immediately instead of waiting
for the next filesystem event.
- HTTP routes mounted at `/api/v0/context` when the manager is up.

### Coderd stub

`coderd/agentapi/context.go` returns `drpcerr.Unimplemented` for
`PushContextState`. The real handler that persists
`workspace_agent_context` rows, hydrates chats, and emits dirty events
lives in CODAGT-569.

## Tests

24 tests across `agent/agentcontext` cover types, paths, resolver
behavior with file caps, skill containers, MCP secret omission, symlink
target validation, the recursive watcher firing on real fsnotify events,
manager source CRUD / `Resync` / `SeedSources` / `Run` lifetime, the
HTTP API, the DRPC adapter, and the push retry / initial-flag /
unimplemented paths. Passes `go test -race -count=2`.

`TestAgent_ContextStatePushed` boots a full agent against
`agenttest.FakeAgentAPI` (which now records `PushContextState` traffic)
and asserts the seeded `AGENTS.md` appears in a snapshot push with
`schema_version = 1`.

<details>
<summary>Notes for reviewers</summary>

- Source CRUD is workspace-agent-token only; coderd is not in the path
for source mutation.
- Per-resource cap 64 KiB, aggregate 2 MiB, count cap 500; resources
past the cap ship with `StatusExcluded` and an empty payload so the
aggregate hash still detects content edits. MCP-emitted resources
enforce both a per-provider count cap and the aggregate byte cap.
- Symlinks inside the scan root are followed; symlinks pointing outside
(or broken) are rejected with `StatusExcluded` so credentials reachable
via a stray symlink stay off the wire.
- The initial push gates `lifecycle = ready` in the eventual full
design. For this PR the `SeedSources` plus `handleManifest`-driven
`Resync` keeps the snapshot fresh; the live push loop ships now and
DRPCPusher translates the coderd `Unimplemented` stub into a clean exit.
- The `PLUGIN`/`HOOK`/`SUBAGENT`/`COMMAND` kinds are reserved in proto
and Go enums but unused; the Claude Code plugin resolver ships in a
follow-up that does not need a schema migration.
- Two follow-ups remain, both tracked by
[CODAGT-569](https://linear.app/codercom/issue/CODAGT-569/enable-agent-api-v210-pushcontextstate-bump-currentminor-wire-coderd):
(1) the chatd-side handler that persists snapshots and dirties chats;
(2) the `coder exp chat context` CLI command set for
`list`/`show`/`add`/`remove`/`refresh`.

</details>

_This PR was authored by Coder Agents on Kyle Carberry's behalf._
2026-06-08 12:08:40 -07:00

399 lines
11 KiB
Go

package agentcontext_test
import (
"context"
"os"
"path/filepath"
"runtime"
"sync/atomic"
"testing"
"time"
"github.com/stretchr/testify/require"
"github.com/coder/coder/v2/agent/agentcontext"
"github.com/coder/coder/v2/testutil"
)
// TestMain points the test binary's HOME (and USERPROFILE on
// Windows) at a fresh empty directory before any test runs.
// The package's built-in scan roots (~/.coder,
// ~/.coder/skills, ~/.claude/plugins/cache) canonicalize
// against this directory, so they resolve to non-existent
// paths and the resolver silently skips them. Without this,
// running the tests on a developer host pulls real Coder and
// Claude config files into snapshots and breaks every
// Len(Resources, N) assertion.
func TestMain(m *testing.M) {
home, err := os.MkdirTemp("", "agentcontext-test-home-")
if err != nil {
panic(err)
}
if err := os.Setenv("HOME", home); err != nil {
panic(err)
}
if runtime.GOOS == "windows" {
if err := os.Setenv("USERPROFILE", home); err != nil {
panic(err)
}
}
code := m.Run()
_ = os.RemoveAll(home)
os.Exit(code)
}
func newTestManager(t *testing.T, opts agentcontext.ManagerOptions) *agentcontext.Manager {
t.Helper()
opts.Logger = testutil.Logger(t).Named("agentcontext-test")
m := agentcontext.NewManager(opts)
t.Cleanup(func() { _ = m.Close() })
return m
}
func TestManager_InitialSnapshotIsPopulated(t *testing.T) {
t.Parallel()
dir := t.TempDir()
mustWriteFile(t, filepath.Join(dir, "AGENTS.md"), "boot snapshot")
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return dir },
})
snap := m.Snapshot()
require.Equal(t, uint64(1), snap.Version)
require.Equal(t, agentcontext.CurrentSchemaVersion, snap.SchemaVersion)
require.Len(t, snap.Resources, 1)
}
func TestManager_AddSourceTriggersResolve(t *testing.T) {
t.Parallel()
wd := t.TempDir()
src := t.TempDir()
mustWriteFile(t, filepath.Join(src, "AGENTS.md"), "from source")
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
AllowedRoots: []string{wd, src},
})
ctx := testutil.Context(t, testutil.WaitLong)
go func() { _ = m.Run(ctx) }()
t.Cleanup(func() { _ = m.Close() })
// Subscribe before mutating so we observe the broadcast.
ch, unsub := m.SubscribeChanges()
defer unsub()
added, err := m.AddSource(agentcontext.Source{Path: src})
require.NoError(t, err)
require.Equal(t, src, added.Path)
select {
case <-ch:
case <-time.After(testutil.WaitShort):
t.Fatalf("expected a change broadcast after AddSource")
}
snap := m.Snapshot()
require.Greater(t, snap.Version, uint64(1))
found := false
for _, r := range snap.Resources {
if r.Kind == agentcontext.KindInstructionFile && r.SourcePath == src {
found = true
}
}
require.True(t, found, "expected AGENTS.md attributed to the user source")
}
func TestManager_AddSourceRejectsOutsideAllowedRoots(t *testing.T) {
t.Parallel()
wd := t.TempDir()
outside := t.TempDir()
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
AllowedRoots: []string{wd},
})
_, err := m.AddSource(agentcontext.Source{Path: outside})
require.Error(t, err)
}
// TestManager_AddSourceAcceptsLateWorkingDir mirrors the agent's
// real boot order: AllowedRoots is configured before the
// manifest provides the workspace working directory. The Manager
// must consult WorkingDir on every check so paths under the
// resolved working dir validate once the manifest lands.
func TestManager_AddSourceAcceptsLateWorkingDir(t *testing.T) {
t.Parallel()
wd := t.TempDir()
var resolved atomic.Pointer[string]
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string {
if p := resolved.Load(); p != nil {
return *p
}
return ""
},
AllowedRoots: []string{"/never-used-home"},
})
// Before the manifest "loads", workingDir is empty; sources
// under wd must be rejected.
_, err := m.AddSource(agentcontext.Source{Path: wd})
require.Error(t, err)
// After the manifest "loads", workingDir resolves and the
// same path validates without restarting the Manager.
resolved.Store(&wd)
_, err = m.AddSource(agentcontext.Source{Path: wd})
require.NoError(t, err)
}
func TestManager_AddSourceIsIdempotent(t *testing.T) {
t.Parallel()
wd := t.TempDir()
src := t.TempDir()
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
AllowedRoots: []string{wd, src},
})
added1, err := m.AddSource(agentcontext.Source{Path: src})
require.NoError(t, err)
added2, err := m.AddSource(agentcontext.Source{Path: src})
require.NoError(t, err)
require.Equal(t, added1.Path, added2.Path)
sources := m.Sources()
require.Len(t, sources, 1)
}
func TestManager_RemoveSource(t *testing.T) {
t.Parallel()
wd := t.TempDir()
src := t.TempDir()
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
AllowedRoots: []string{wd, src},
})
_, err := m.AddSource(agentcontext.Source{Path: src})
require.NoError(t, err)
require.NoError(t, m.RemoveSource(src))
require.Empty(t, m.Sources())
err = m.RemoveSource(src)
require.ErrorIs(t, err, agentcontext.ErrSourceNotFound)
}
func TestManager_HasSource(t *testing.T) {
t.Parallel()
wd := t.TempDir()
src := t.TempDir()
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
AllowedRoots: []string{wd, src},
})
canonical, ok := m.HasSource(src)
require.False(t, ok)
require.Equal(t, src, canonical)
_, err := m.AddSource(agentcontext.Source{Path: src})
require.NoError(t, err)
canonical, ok = m.HasSource(src)
require.True(t, ok)
require.Equal(t, src, canonical)
}
func TestManager_ResyncReturnsLatestSnapshot(t *testing.T) {
t.Parallel()
wd := t.TempDir()
mustWriteFile(t, filepath.Join(wd, "AGENTS.md"), "first")
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
})
ctx := testutil.Context(t, testutil.WaitLong)
runDone := make(chan struct{})
go func() {
defer close(runDone)
_ = m.Run(ctx)
}()
t.Cleanup(func() {
_ = m.Close()
<-runDone
})
// Mutate AGENTS.md and call Resync. The returned
// snapshot must reflect the new content.
require.NoError(t, os.WriteFile(filepath.Join(wd, "AGENTS.md"), []byte("second content edit"), 0o600))
snap, err := m.Resync(ctx)
require.NoError(t, err)
require.Len(t, snap.Resources, 1)
require.Equal(t, "second content edit", string(snap.Resources[0].Payload))
}
// TestManager_ResyncCanceledKeepsLiveSnapshot guards CRF-44:
// a context cancellation mid-walk must not replace the live
// Snapshot with an empty one. Resync returns the existing
// Snapshot and ctx.Err() instead of publishing a stub.
func TestManager_ResyncCanceledKeepsLiveSnapshot(t *testing.T) {
t.Parallel()
wd := t.TempDir()
mustWriteFile(t, filepath.Join(wd, "AGENTS.md"), "live content")
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
})
// Capture the live snapshot the Manager populated at
// construction time.
live := m.Snapshot()
require.Len(t, live.Resources, 1)
require.Equal(t, "live content", string(live.Resources[0].Payload))
// Cancel the context before calling Resync so
// ResolveContext observes the cancellation.
ctx, cancel := context.WithCancel(context.Background())
cancel()
snap, err := m.Resync(ctx)
require.ErrorIs(t, err, context.Canceled)
// The returned snapshot must still expose the live
// resources, not an empty result from the canceled walk.
require.Len(t, snap.Resources, 1)
require.Equal(t, "live content", string(snap.Resources[0].Payload))
// The next Snapshot call must also return live content;
// no stub was published.
after := m.Snapshot()
require.Equal(t, live.Version, after.Version)
require.Len(t, after.Resources, 1)
}
func TestManager_InitialSourcesSeeded(t *testing.T) {
t.Parallel()
wd := t.TempDir()
src := t.TempDir()
mustWriteFile(t, filepath.Join(src, "AGENTS.md"), "from initial")
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
AllowedRoots: []string{wd, src},
InitialSources: []agentcontext.Source{{Path: src}},
})
sources := m.Sources()
require.Len(t, sources, 1)
require.Equal(t, src, sources[0].Path)
snap := m.Snapshot()
require.Len(t, snap.Resources, 1)
require.Equal(t, src, snap.Resources[0].SourcePath)
}
// TestManager_SeedSourcesLateBindsAfterManifest models the
// agent's behavior when CODER_AGENT_EXP_*_DIRS contains a
// relative path that cannot resolve until the manifest's
// working directory lands. SeedSources must adopt the
// previously-unresolvable path, bypass AllowedRoots
// validation, and trigger a re-resolve.
func TestManager_SeedSourcesLateBindsAfterManifest(t *testing.T) {
t.Parallel()
wd := t.TempDir()
late := t.TempDir()
mustWriteFile(t, filepath.Join(late, "AGENTS.md"), "late binding")
// AllowedRoots intentionally omits `late` so AddSource
// would reject it. SeedSources must accept it anyway,
// since the path comes from the trusted template config.
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
AllowedRoots: []string{wd},
})
require.Empty(t, m.Sources())
m.SeedSources([]agentcontext.Source{{Path: late}})
sources := m.Sources()
require.Len(t, sources, 1)
require.Equal(t, late, sources[0].Path)
snap, err := m.Resync(testutil.Context(t, testutil.WaitShort))
require.NoError(t, err)
require.Len(t, snap.Resources, 1)
require.Equal(t, late, snap.Resources[0].SourcePath)
}
func TestManager_CloseIsIdempotent(t *testing.T) {
t.Parallel()
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return t.TempDir() },
})
require.NoError(t, m.Close())
require.NoError(t, m.Close())
}
func TestManager_RunOnce(t *testing.T) {
t.Parallel()
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return t.TempDir() },
})
ctx, cancel := context.WithCancel(testutil.Context(t, testutil.WaitShort))
defer cancel()
go func() { _ = m.Run(ctx) }()
// Wait for Run to claim the running flag, then verify the
// second call rejects with a deterministic error rather than
// racing the scheduler.
select {
case <-agentcontext.ManagerStarted(m):
case <-ctx.Done():
t.Fatalf("manager never started: %v", ctx.Err())
}
err := m.Run(ctx)
require.Error(t, err)
require.Contains(t, err.Error(), "more than once")
cancel()
_ = m.Close()
}
func TestManager_SubscribeBroadcastOnChange(t *testing.T) {
t.Parallel()
wd := t.TempDir()
src := t.TempDir()
m := newTestManager(t, agentcontext.ManagerOptions{
WorkingDir: func() string { return wd },
AllowedRoots: []string{wd, src},
})
ctx := testutil.Context(t, testutil.WaitLong)
go func() { _ = m.Run(ctx) }()
ch, unsub := m.SubscribeChanges()
defer unsub()
_, err := m.AddSource(agentcontext.Source{Path: src})
require.NoError(t, err)
select {
case <-ch:
case <-time.After(testutil.WaitShort):
t.Fatal("expected subscriber to be notified")
}
}